Propuesta de modificación

The purpose of this quantitative cross-sectional survey using secondary data analysis was to investigate the association between health-system characteristics and access to maternal health medicines. The study participants included 37 low and middle-income countries who participated in USAID MCHIP and WHO pharmaceutical 2011/2012 survey. I used the IBM SPSS Statistics 21 software to answer the questions and test the hypotheses listed below:

Question 1: Is there a significant association between governance and access to essential maternal health medicines in low and middle-income countries?

Ho1: there is no significant association between governance and access to essential maternal health medicines in low and middle-income countries?

Ha1: there is a significant association between the governance and access to essential maternal health medicines in low and middle-income countries?

Question 2: Is there a significant association between pharmaceutical supply and access to essential maternal health medicines in low and middle-income countries?

Ho2: there is no significant association between pharmaceutical supply and access to essential maternal health medicines in low and middle-income countries?

Ha2: there is a significant association between pharmaceutical supply and access to essential maternal health medicines in low and middle-income countries?

Question 3: Is there a significant association between the quality of health facility and access to essential maternal health medicines in low and middle-income countries?

Ho3: there is no significant association between the quality of health facility and access to essential maternal health medicines in low and middle-income countries?

Ha3: there is a significant association between the quality of health facility and access to essential maternal health medicines in low and middle-income countries?

Question 4: Is there a significant association between quality of service delivery and access to essential maternal health medicines in low and middle-income countries?

Ho4: there is no significant association between quality of service delivery and access to essential maternal health medicines in low and middle-income countries?

Ha4: there is a significant association between quality of service delivery and access to essential maternal health medicines in low and middle-income countries?

Question 5: Is there a significant relationship between health financing and access to essential maternal health medicines in low and middle-income countries?

Ho5: there is no significant association between health financing and access to essential maternal health medicines in low and middle-income countries?

Ha5: there is a significant association between health financing and access to essential maternal health medicines in low and middle-income countries?

Data Collection

Data set for both surveys are publicly available and can be found on the website of WHO and USAID MCHIP. There was no discrepancy between the plan presented in Chapter 3 and the actual data collection.

Primary data for the MCHIP survey was coordinated by the MCHIP maternal health team in Washington, D.C., during the months of January, February and March 2012. Contact information was compiled for an identified focal person in each of the 43 countries initially targeted by the survey. An identified coordinator for each country was sent an e-mail with anticipated dates and activities six weeks in advance of receiving the survey. He or she was instructed to contact national counterparts in the government as well as leading implementing partners. The country coordinator was given a timeline of pending requests and asked to arrange meetings with national consultative groups to ensure a national participatory process for the completion of the survey instruments. In most cases this was possible. Key stakeholders from government, ministries, MCHIP programs, other USAID bilateral programs, UN partners and other implementing agencies met to collect data and respond to the 46-item questionnaire and the scale-up map. In most cases, these consultative groups found it necessary to meet twice to ensure accuracy and completeness of responses.

The WHO pharmaceutical country survey data was collected from all 193 member states using a user-friendly electronic questionnaire that included comprehensive instruction manual and glossary. Countries were requested to enter the results from previous surveys and to provide centrally available information compiled data comes from international sources (e.g. the World Health Statistics), surveys conducted in the previous years and country level information collected in 2011.

Inclusion and Exclusion Criteria

The original MCHIP survey dataset included records of 37 countries from Africa, Asia and the Americas; while those of WHO country surveys included records for 193 member countries. The records were assessed for eligibility based on the inclusion and exclusion criteria.

Countries were included if they participated in the USAIDS MCHIP survey. They were excluded if they did not participate in the MCHIP survey.

Review of Statistical Assumptions

The study analysis included standard multiple regressions on one outcome variable (Access to live-saving maternal health medicines) and six independent variables (Strength of health system governance, pharmaceutical procurement and distribution, quality of health services and facilities, health care financing, and reporting of relevant maternal health medicine indicators). Responses to forty-four numeric and string question items that best represented these variables were collected and grouped to form composite variables for the regression analysis. I reviewed the key assumptions of logistic regression analysis, in particular, the magnitude of missing data, presence of multi-collinearity and outliers, and the compliance with a minimum of 10 cases per variable category.

Outliers. Using Z-score method to identify outliers, analysis showed that data were positively skewed in four numeric indicators (number of hospital beds per 10,000; pharmacist per 10,000pop; and physicians per 10,000 population and Nurses and midwives per 10,000 population) with skewness ranging of 4.148, 3.154, 4.891, 2.910 respectively. To make data normally distributed and suitable for further analysis, I applied a two-step process. Data were

first ranked using SPSS transform function. They were then transformed and computed to form new variables using SPSS Transform and IDF normal function.

Missing data. The first part of the missing data analysis showed that 17 of the 44 indicators for the independent and dependent variables included in the study had some missing data. The top five indicators were: public insurance coverage for medicines on EML (43.2%), Access to essential Medicines as part of fulfilment of the right to health (43.2%), Total health expenditure as percent of GDP (37.8%), availability of national guidelines for good distribution practices (37.8%), and presence of public sector procurement policies (35.1%).

Figure 2. Summary of missing data in the study database. Generated from SPSS multiple imputation analysis of missing data patterns.

High rates of missing data can introduce bias and compromise the validity of the study findings. Analysis of missing value patterns suggested some monolithic pattern of missing values across some variables. Multiple imputations, one of the most recommended missing data handling techniques, were used to replace the missing values with five imputations based on the predictive values derived from the observed data (Sterne et al., 2009).

Multicollinearity. Another test was a bivariate correlation analysis of the 12 independent variables in SPSS using the Pearson correlation coefficient. The results showed no value of the Pearson correlation coefficient equal or above .80, confirming that there was no multicollinearity between the variables of interest. All the correlation coefficients were below .60. The second test was the collinearity diagnostics in linear regression that examined the tolerance rate and variance inflation factor (VIF) (Field, 2013). The results found no tolerance rate below .10 or VIF above 10. All the tolerance values were above .70 and VIF below two, indicating that there was no multicollinearity.

Sample Size and Minimum Number of Cases in Each Variable Category.

As a rule of thumb, each category of the independent variables included in logistic regression analysis must have a minimum of 10 cases (Vittinghoff & McCulloch, 2007). In that regard, the total sample of 37 was large enough to achieve sufficient statistical power. Some categories of the independent variables were combined to satisfy the rule of 10. The lowest cell count was 27 cases. All the other categories had more than 27 cases.

Variable Categorization and Coding

The analysis included one outcome and six independent variables. Responses to forty-four numeric and string question items that best represented these variables were collected and combined to form composite dichotomous variables for the regression analysis.

The outcome variable, Access to Medicines had three variables (coded 1. Yes, 2. No):

availability of medicines, affordability of medicines, accessibility of medicines. Countries with less than 50% of the combined scores were assumed to have poor access (coded: 0,) while those with above 50% of combined score, good access (desired outcome, coded 1).

Similar procedures were used to organize the six independent variables. Strength of Health System Governance (coded 1 strong health system governance, 2. Weak health system governance) comprised of three items: medicine is approved at national level, guidelines for treatment are available, access to medicines and technologies are recognized as part of the right to health. Health system procurement, distribution and supply variable (coded, 1. Strong, 2.

Weak) comprised of two items: medicines procurement system, medicines distribution and supply system. Quality of health facility variable (coded 1. High quality, 2. Poor quality) had one item: Number of hospital beds per 1000 population. Similarly, the quality of health services variable (coded 1. High quality, 2. Poor quality) had three items (adequacy of human resources for health, comprehensiveness or education and trainings for human resources for health, Midwife and Birth attendants scope of practice. Health system financing (coded 1. Robust 2. Not robust) had three items: Total health expenditure as % of GDP, Total Pharmaceutical, expenditure as % of health expenditure, private health expenditure as % of total health expenditure. Finally, the sixed variable – reporting of key maternal health medicines (coded 1.

Strong reporting and 2. reporting had a single item: national reporting on key maternal medicines indicators

Descriptive Analysis

About 53% of countries enrolled in the study reported relatively good access to essential maternal health medicines compared to 47% that reported poor access – based on three indicators: availability, affordability, and accessibility of essential maternal health medicines.

Medicines were often (75% of the time) available in only 24% of countries, affordable in 56% of

countries and accessible in 13.5% of countries. Oxytocin was more readily available (75% of the time) in 44% of countries compared with MgSO4 in 37% of countries.

Regarding health systems governance, more countries (61.7%) recognize access to medicines as a right to health in their constitution compared with those that had maternal health medicines on their EML (52.8%) and had guidelines treatment (49.3%). However, amongst all three indicators for strength of health system governance, availability of medicines on the EML seemed to have a statistically significant relationship with access to medicines (p<0.001).

Countries seemed to have better medicine procurement systems than distribution systems.

More countries (62%) had strong procurement systems than they did strong distribution systems (and 44% respectively). Independently cross-tabbed, neither of these indicators showed statistically significant association with the outcome variable. In terms of quality of health facilities, only 24% of countries reported relatively substantial infrastructures for health and this indicator had a statistically significant association with access to essential maternal health medicines (p <0.001).

Results differed slightly with indicators for quality of health services. Comparatively, their appeared to be the most gap in midwives and birth attendants scope of practice among countries. Nearly half (45%) of countries reported less comprehensive education and trainings for human resources for health while a little more than half (59% and 67% respectively) reported relatively less adequate human resources for health and narrow scope of practice for midwives and birth attendants. Except for education and training, other indicators for quality of health services showed statistically significant associations with the outcome variable. Few countries (14% and 35% respectively) appeared to have robust health system financing and report key

maternal health medicines indicators. These indicators showed statistically significant association with access to essential maternal health medicines.

Table 4

Treatment guidelines 37(100) 18(49) 19(50) 26.260 0.000 Access to Medicine

Infrastructure. 37(100) 24(65) 13(34) 14.300 0.000 Quality of Health

Statistical Analysis Findings by Research Questions and Hypotheses

The analysis report below includes the outcomes of bivariate and logistic regression analyses, based on a sample of 37 resource poor countries in Africa, Asia, and America who met the inclusion criteria. The first step was a bivariate (Chi-square) analysis of the association between six independent variables (Health system Governance, medicine procurement, distribution and supply, quality of health facilities, quality of health services, health financing and data reporting for key maternal health medicines.) and Access to essential maternal health medicines. Findings showed that all the independent variables, with the exception of quality of health services had a statistically significant association with access to maternal health medicines and achieved a p-value < .05. The second step included a multiple logistic regression analysis to test which of the five independent variables that had a statistically significant association with the outcome variable and a p-value < .05. significantly predicted access to essential maternal health medicines. The results of the regression indicated that three predictors explained 27% of the variance (R2 =.266, F(5,162)=13.12, p<.01). It was found that the strength of a country’s medicine procurement, distribution, and supply system significantly predicted access to essential maternal health medicines (β= .41, p<.001), as did robustness or health system financing (β= -.51, p<.001), and quality of health facilities (β= -.34, p<.05).

Research question 1. The first question was stated as follows: Can access to live-saving maternal health medicines be predicted based on the strength of a country’s health system governance?

Ho1: Access to essential maternal health medicines cannot be predicted based on the strength of a country’s health system governance?

Ha1: Access to essential maternal health medicines can be predicted based on the strength of a country’s health system governance?

A chi-square test of independence was performed to examine the relation between strength of health system governance and access to live-saving maternal health medicines. The vast majority (71.8%) of countries enrolled in the study had weak health system governance as compared to 28.2% with strong health system governance.

The standardized deviations to measure the magnitude of the difference between observed and expected values on women’s access to essential maternal health medicines by the strength of a country’s health system governance (Spiegel, Schiller, and Srinivasan, 2009) showed that stronger health system governance had a negative deviation towards poor access to essential maternal health medicines. The inverse direction was observed for relatively weaker health system governance (table 6). A negative standardized deviation indicates that the observed values are lower than expected and a positive standardized deviation that the observed values are higher than expected. More countries (28%) with weaker health system governance had poor access to essential maternal health medicines compared with those with strong health system governance (11%).

Table 5

Standardized Deviations of Health System Governance by Access to Maternal Health Medicines Access to Maternal Health Medicines

Good Poor

Health Systems Governance

Strong +1.6 -3.3

Weak -1.0 +2.1

The bivariate analysis showed a statistically significant association between health system governance and access to essential maternal health medicines χ(1) = 18.999, OR = 1.368, CI [ 1.239, 1.150], p = .000. This finding may suggest that women in countries with stronger health system governance were more likely to have better access to maternal health medicines than those in countries with weaker health system governance. However, a multivariate analysis that followed did not retain the statistically significant association between the strength of health system governance and access to essential maternal health medicines (β= -.160, t = -1.351, CI [ -0.394, 0.074], p =.178). Therefore, I accept the null hypothesis and conclude that access to essential maternal health medicines cannot be predicted based on the strength of a country’s health system governance.

Research question 2. The second question was stated as follows: Can access to essential maternal health medicines be predicted based on the strength of a country’s medicine procurement, distribution, and supply system?

Ho2: Access to essential maternal health medicines cannot be predicted based on the strength of a country’s medicine procurement, distribution, and supply system?

Ha2: Access to essential maternal health medicines can be predicted based on the strength of a country’s medicine procurement, distribution, and supply system?

A chi-square test of independence was performed to examine the relation between strength of Medicine procurement, distribution and supply system and access to essential maternal health medicines. The vast majority (64.3%) of countries enrolled in the study had weak medicine procurement, distribution and supply system compared to 35.7% with strong medicine procurement, distribution and supply system.

The standardized deviations to measure the magnitude of the difference between observed and expected values on women’s access to essential maternal health medicines by the strength of a country’s health system governance (Spiegel, Schiller, and Srinivasan, 2009) showed that stronger medicine procurement, distribution and supply system had a negative deviation towards poor access to essential maternal health medicines (table 6). A negative standardized deviation indicates that the observed values are lower than expected and a positive standardized deviation that the observed values are higher than expected.

More countries (24.8%) with weaker medicine procurement, distribution and supply had poor access to essential maternal health medicines compared with those with strong Medicine procurement, distribution and supply system (8.1%).

Table 6

Standardized Deviations of Medicine procurement, distribution and supply by Access to Maternal Health Medicines

The bivariate analysis showed a statistically significant association between medicine procurement, distribution and supply system and access to essential maternal health medicines χ(1) = 8.676, OR = 3.740, CI [ 1.486, 9.411], p = .003.

A multivariate analysis that followed retained the statistical significant association between medicine procurement, distribution and supply system of a country and access to essential maternal health medicines (β= -.411, t = - 3.641, CI [ - 0.634, -.188], p <0.001). These

findings suggest that women in countries with stronger health medicine procurement, distribution and supply systems were more than 3 times more likely to have better access to maternal health medicines than those in countries with weaker medicine procurement, distribution and supply system. Therefore, I reject the null hypothesis and conclude that access to essential maternal health medicines can be predicted based on the strength of a country’s medicine procurement, distribution and supply system.

Research question 3. The third question was stated as follows: Can access to essential maternal health medicines be predicted based on the quality of health facilities in a country?

Ho3: Access to essential maternal health medicines cannot be predicted based on the quality of health facilities in a country

Ha3: Access to essential maternal health medicines can be predicted based on the quality of health facilities in a country.

A chi-square test of independence was performed to examine the relation between quality of health facility and access to live-saving maternal health medicines. The vast majority (65.6%) of countries enrolled in the study had relatively good quality health facilities compared to 34.4%

with poor quality.

The standardized deviations to measure the magnitude of the difference between observed and expected values on women’s access to essential maternal health medicines by the strength of a country’s health system governance (Spiegel, Schiller, and Srinivasan, 2009) showed that countries with good quality health facilities had a negative deviation towards poor access to essential maternal health medicines (table 6). The reverse was the case for countries with poor quality health facilities. A negative standardized deviation indicates that the observed

values are lower than expected and a positive standardized deviation that the observed values are higher than expected.

More countries (33.7%) with poor quality health facilities had poor access to essential maternal health medicines compared with those with higher quality health facilities (11.7%).

Table 7

Standardized Deviations of Quality of Health Facilities by Access to Maternal Health Medicines Access to Maternal Health Medicines

Good Poor

Quality of Health Facilities

Strong +1.0 -2.0

Weak -1.3 +2.8

The bivariate analysis showed a statistically significant association between quality of health facilities and access to essential maternal health medicines χ (1) = 14.600, OR = 3.781, CI [ 1.849, 7.733], p <0.001. A multivariate analysis that followed retained the statistical significant association between the quality of health facilities in a country and access to essential maternal

The bivariate analysis showed a statistically significant association between quality of health facilities and access to essential maternal health medicines χ (1) = 14.600, OR = 3.781, CI [ 1.849, 7.733], p <0.001. A multivariate analysis that followed retained the statistical significant association between the quality of health facilities in a country and access to essential maternal